← Reference library

REFERENCE LIBRARY

Inhibitory InterneuronsInhibitory Interneurons

Roughly one in five cortical neurons sends nothing to distant regions and only pulls a local switch. They are not an off button for signals but the hand that shapes and times them — cortical rhythms, the gate of attention, and whether a circuit runs away from itself are all in their charge.

A minority that only works locally

Cortical neurons split roughly in two. Excitatory cells (mostly pyramidal cells) make up around 80%, use glutamate as their messenger, and often send axons to far-away regions — they are the ones doing the talking. The remaining ten to twenty percent are interneurons: they use GABA, whose effect is to make the target harder to fire, and their axons essentially never leave the neighbourhood they wire.

Don't let "minority" mislead you. They come in far more varieties than excitatory cells, and each variety has its own landing site — where it attaches on the target cell determines what it can do. The three classes defined by molecular markers together cover the great majority of cortical interneurons:

PV (parvalbumin) · basket cells

They wrap the target's cell body and its output stalk like a basket — right where the action potential is born, so pressing there is a clean cut. They themselves fire fast and precisely and barely fatigue. The beat of cortical gamma is kept mainly by these cells.

SST (somatostatin) · Martinotti cells

Their axons climb to the far tips of the pyramidal cell's apical dendrite, exactly where top-down feedback arrives. So what they suppress isn't "this cell must not fire" but "this particular stream of input must not be heard" — a hand that picks among inputs, not a master switch.

VIP (vasoactive intestinal peptide)

The most interesting class: they mainly inhibit SST cells. Inhibiting the inhibitor nets out as release. VIP cells become active when an animal starts running, gets rewarded, or begins attending — they are the switch that temporarily opens a patch of cortex.

E-I balance

Excitation and inhibition are a seesaw tied firmly together: when excitation rises, inhibition rises right behind it. That isn't waste — it keeps the circuit from igniting out of control (that is epilepsy) while preserving an adjustable time window.

Inhibition doesn't switch things off — it shapes and times them

In time: a population pressed down together and released together turns continuous time into alternating stretches of "can speak" and "speaking is pointless." That is where brain oscillations come from, and why when a signal arrives can matter more than how much of it arrives.

In space: local inhibition lets the most active point push down its surround, pulling weak contrasts apart — the retina and the cortex use the same trick to sharpen edges.

As a switch: since inhibition is a standing background, lifting it becomes an extremely fast way to grant passage (the VIP → SST → pyramidal chain). Attention, motivation and locomotor state often modulate cortex along this route rather than by adding excitation directly.

When it goes wrong

Epilepsy is the most direct imbalance: the inhibitory side fails to hold, a huge population is dragged onto a single beat, and at maximal synchrony function is lost rather than gained.

In schizophrenia, reduced markers of prefrontal PV cells and weakened task-related gamma have been observed repeatedly, and the lead is taken seriously (→ prefrontal cortex). But state it plainly: this is largely post-mortem and correlational evidence, and the step from "PV cells are affected" to "PV cells cause the symptoms" has not been causally established. Likewise the "E/I imbalance" account of autism remains a hypothesis, not a conclusion.